Filter Sound from Data

I have frequency-based data sets (with column 1 for time and column 2 for amplitude), and I use the following code to filter out the sound. I'm not convinced that this is the most effective system as I believe this process changes the time at which other signals occur. Is there a better system for filtering data?

BP = [0.8725 0.8925]; %Frequency filter band

Fs = 1/(data(2,1)-data(1,1)); %Sampling Freq.

Y00 = bandpass(data(:,2),BP,Fs); %Filtered data

Answer 1

[Jan]

filtfilt works without shifting the data.

[Sam Hurrell]

How is filtfilt different to Bandpass?

Answer 2

[Star Strider]

That appears to me to be correct.

If you are using the same filter for all the data sets, return the digital filter object after the first set:

[Y00,df] = bandpass(data(:,2),BP,Fs, 'ImpulseResponse','iir'); %Filtered data

to design an elliptic filter (see ellip for details) and return it as well.

Then in subsequent data sets, use it with filtfilt to do the filtering:

Y01 = filtfilt(df, data2(:,2));

That elimiinates the computational expense of designing the same filter for each signal.

.

[Sam Hurrell]

Apologies as I am not well versed in Matlab, but what dyou mean by "return the digital filter", and how does this work with filtfilt?

[Star Strider]

The bandpass and related functions optionally return the digital filter object they created as the second output. See for example the Bandpass Filter Steepness section of the documentation.

data = randn(500,2);                                                % Create Matrix
data(:,1) = linspace(0, 499, 500)'/100;
data2 = randn(500,2);                                               % Create Matrix
t = data(:,1);

BP = [0.8725 0.8925];                                               %Frequency filter band
Fs = 1/(data(2,1)-data(1,1));                                       %Sampling Freq.
[Y00, df] = bandpass(data(:,2),BP,Fs,'ImpulseResponse','iir');      %Filtered data
Y01 = filtfilt(df, data2(:,2));

figure
plot(t, Y00)
grid

figure
plot(t, Y01)
grid

The data are synthetic, however the code demonstrates the correct procedure.

.